Dashboard Overview

Interactive demand analytics powered by Plotly Dash

The Demand Forecasting Dashboard is a Plotly Dash application designed for interactive exploration of prediction data. Unlike static reports, it allows stakeholders to slice data by seasonality and date ranges in real time.

​

Technology Stack

The UI is built using a Python-native stack, ensuring seamless integration with the data science workflow:

• Framework: Plotly Dash
• Components: dash-bootstrap-components (LUMEN theme)
• Data Source: Direct Parquet ingestion (High Performance)
• Styling: Custom CSS overlay on Bootstrap Grid

​

Key Features

The dashboard (app-ui/app.py) focuses on three core analytical views:

​

1. Global KPIs

Located at the top, providing immediate context:

• Total Rentals: Aggregated demand for the selected period.
• Avg Temp: Normalized temperature context.
• Registered Users: Growth metric for user base.

​

2. Seasonality Analysis

Visualization Logic

# How we visualize filtered seasonality
fig2 = px.box(
    filtered_df, 
    x='season_label', 
    y='cnt', 
    color='season_label',
    category_orders={'season_label': ['Spring', 'Summer', 'Fall', 'Winter']},
    color_discrete_map={
        'Spring': '#66BB6A', 
        'Summer': '#FDD835', 
        'Fall': '#FF7043', 
        'Winter': '#42A5F5'
    }
)

This box plot reveals the distribution of demand across seasons, highlighting outliers (e.g., unusually high demand in Winter).

​

3. Temperature Correlation

A scatter plot correlating normalized temperature (temp) with demand (cnt). This validates the model’s reliance on weather features:

• Positive Correlation: Higher temps generally lead to higher demand.
• Color Coding: Points are colored by season to show clustering.

​

Architecture

The Dash app runs as a standalone Docker service (app-ui), decoupled from the heavy ML training pipelines.

It reads the same processed parquet files that the model uses, ensuring 100% data consistency between training and visualization.